The isolation of novel genes and polypeptides from semi-random sequences is currently limited by the need to screen a large, genetically diverse population of cells in order to obtain the sequence(s) of interest. For example, a polypeptide string of 10 amino acids has 20.sup.10 or approximately 10.sup.13 possible permutations. If 10 of these permutations had a desirable characteristic (such as the ability to bind a specific antigen), then a population of 10.sup.12 would have to be screened for the expectation of finding one desirable novel gene. Through the use of conventional methods (expressing novel genes via microorganisms), the screening of a large number of new sequences for a specific property is virtually unfeasible, unless the novel gene provides the organism with a distinct growth or survival advantage. Indeed, under the current state of the art, the 10.sup.12 independently transformed microorganisms would have to be screened individually to locate that one desirable novel gene.
Within present screening procedures for detecting novel gene products which are localized within cells, colonies derived from each transformed cell must be treated to break open the cells. Typically 1000-2000 bacterial colonies per standard petri dish are lysed (e.g., by chloroform) for the screening procedure. Thus, to examine 10.sup.12 transformed organisms, 500,000 to 1 billion petri dishes would be necessary. In addition, 10,000 to 100,000 liters of logarithmically dividing cells may be necessary for producing the large numbers of transformable cells.
Alternatively, where a gene product is secreted and attached to the outside of a cell, it may be detected by its ability to bind a fluorescent compound or other marker. In these cases, cell sorters may be used to screen for the synthesis of a novel desirable polypeptide. However, even at a flow rate of 5,000 cells per second, it would take a cell sorter over 60 years to screen 10.sup.12 cells. Thus, present day screening methods which are both extremely costly and time-consuming, effectively prohibit the isolation of novel genes and polypeptide from semi-random sequences.
In addition to the methods briefly discussed above, Fields and Song (Nature ;340:245-246, 1989) proposed a method for selectably obtaining polypeptides which specifically bind to other polypeptides, using the domains of the yeast GAL4 gene. However, this system has serious limitations. First, only polypeptide-polypeptide binding may be selected; polypeptide-nonpolypeptide interactions are excluded. Second, both the known and novel binding polypeptides have to be expressed in yeast at reasonably high levels and in "native" conformations for the method to have commercial applicability. Third, glycosylated polypeptides or polypeptides that have special modifications may also be excluded by this method. Fourth, it is not clear whether random or semi-random sequences can work, given that they used known polypeptides whose physical interactions were well-established and yet showed only 4.5% of the control GAL4 activity. Fifth, Fields and Song used very large sequences: 633 amino acids of the SNF1 protein and 322 amino acids of the SNF4 protein, which have evolved secondary structures that interact with each other. Sixth, using their method for semi-random sequences of even 10.sup.10 diversity obviates the need for extremely large amounts of DNA, modifying enzymes, and competent yeast cells.
Contrary to previously disclosed methods, the present invention describes a method for cell-free screening of novel genes and polypeptides. This method avoids the problems associated with large numbers of transformed organisms as well as the limitations of the method disclosed by Fields and Song, and may be completed within a few weeks. Therefore, the methodology allows a substantial time and monetary saving in the isolation of novel gene products.